Deep learning for multimedia application

Abstract

Recent advancements in speech applications prominently feature Deep Learning, driving significant progress in the challenging task of separating speech signals from multi-speaker speech mixtures. Speech Separation models have a wide range of applications ranging from enhancing the performance of hearing aids, use in telecommunications and serving as a pre-processing model in automatic speech recognition. In the following report, we analyze recent advancements in Deep Learning models for Monaural Speech Separation and discuss some ideas for the future direction of this work.